From: loic d'Anterroches (loa_at_kt.dtu.dk)
Date: Thu 04 Mar 2004 - 09:57:40 GMT
Hi,
> Yes, you are right: heartbeat doesn't need to manage the ip takeover,
> vserver does. We emit an unsolitcited arp reply in the vserver pre-start
> script in order to inform all hosts in the local subnet about the
> takeover.
> We don't run the rebootmgr because we noticed that its open socket
> in the vserver filesystem prevented heartbeat from unmounting the
> vserver filesystem and the takeover hangs.
>
> Our heartbeat resources are datadisk, vserver and MailTo, where the
> vserver resource script is just a symbolic link to the standard vserver
> script.
Thanks for this confirmation. It is more clear for me. I will also this
rebootmgr issue.
>> Off-topic: If you are using drbd, how are you managing the synchro when
>> the master BA comes online again?
>
> Don't know what you want to know here. We have heartbeat configured with
> "auto_failback off", so when BA comes up again, the vserver stay on BB
> until we interactively force the takeover. So we can have a look at BA
> and eventually start drbd on BA if that doesn't happen automagically.
> If you want to force a full sync, you have to delete all files in
> /var/lib/drbd before reconnecting BA. After drbd has synced the
> disk(s) you can do a "/usr/lib/heartbeat/hb_standby foreign" on BB
> to force the failback of the vservers.
Sorry, my question was not correctly formulated, but you gave me the
answer anyway :-) I was not sure about the synchro back from BB to BA when
BA comes online again, but definitely it looks like doing that manually is
better at least to correctly check BA.
Thanks for your complete and fast answer,
loic
_______________________________________________
Vserver mailing list
Vserver_at_list.linux-vserver.org
http://list.linux-vserver.org/mailman/listinfo/vserver